Greedy bases for Besov spaces (Q717121): Difference between revisions

Let \((e_i)\) be a basis for a Banach space \(X\). One is interested in the best \(n\)-term approximation of \(x\in X\) by some sum \(\sum_{i\in A} a_i e_i\), with \(\text{card}(A)=n\). If this is achieved, up to a constant \(C\) that is independent of \(n\), by picking a set of indices containing \(n\) largest coefficients \(|e_i^*(e_i)|\), then the basis is called greedy, more precisely \(C\)-greedy. This notion was introduced by \textit{S.~V.~Konyagin} and \textit{V.~N.~Temlyakov} in [East J. Approx.\ 5, No.~3, 365--379 (1999; Zbl 1084.46509)]. The main result of the paper says that for \(1<q<\infty\) and \(1\leq p\leq \infty\) the \(\ell_q\)-sum of the \(\ell_p^n\)-spaces, \(X=(\bigoplus_{n=1}^\infty \ell_p^n)_{\ell_q}\), has a greedy basis; a \(1\)-greedy basis exists if and only if \(p=q\). In the limiting cases \(q=1\) and \(p>1\) or \(q=\infty\) and \(p<\infty\) (then \(\ell_q\) should be replaced by \(c_0\)), no greedy bases exist. The proof relies on the characterisation of greedy bases as unconditional and ``democratic'' bases by Konyagin and Temlyakov, the isomorphism of \(X\) with \(Y= (\bigoplus_{N=1}^\infty \ell_q(\ell_p^N))_{\ell_q}\) that depends on the Pełczyński decomposition method, and a technical construction of certain finite unconditional basic sequences in \(\ell_q(\ell_p^N)\). The actual construction is then performed in~\(Y\). Since certain Besov spaces \(B^{\alpha,m}_{p,q}\) on the unit interval are known to be isomorphic to \(X\) and hence to \(Y\), the main result implies the existence of greedy bases in those Besov spaces. Since the isomorphism depends on the decomposition method, it is not canonical, and hence it remains open to find a greedy basis in \(B^{\alpha,m}_{p,q}\) formed by classical function systems.